##// END OF EJS Templates
Merge pull request #4542 from ivanov/history-clear...
Merge pull request #4542 from ivanov/history-clear new `ipython history clear` subcommand This new subcommand is equivalent to ipython history trim --HistoryTrim.keep=0, which is now supported (and also has a more convenient --keep alias) This command comes with a fix to bash completion of flags for ipython history subcmd also included is a tiny fix for nbconvert completion that was pointed out by @jakobgager in #4528 also included is a change to our utils.io.ask_yes_no function that allows for specifying the behavior of a KeyboardInterrupt.

File last commit:

r12409:d918b135
r13694:24e10561 merge
Show More
citation.py
72 lines | 2.1 KiB | text/x-python | PythonLexer
"""Citation handling for LaTeX output."""
#-----------------------------------------------------------------------------
# Copyright (c) 2013, the IPython Development Team.
#
# Distributed under the terms of the Modified BSD License.
#
# The full license is in the file COPYING.txt, distributed with this software.
#-----------------------------------------------------------------------------
#-----------------------------------------------------------------------------
# Code
#-----------------------------------------------------------------------------
__all__ = ['citation2latex']
def citation2latex(s):
"""Parse citations in Markdown cells.
This looks for HTML tags having a data attribute names `data-cite`
and replaces it by the call to LaTeX cite command. The tranformation
looks like this:
`<cite data-cite="granger">(Granger, 2013)</cite>`
Becomes
`\\cite{granger}`
Any HTML tag can be used, which allows the citations to be formatted
in HTML in any manner.
"""
try:
from lxml import html
except ImportError:
return s
tree = html.fragment_fromstring(s, create_parent='div')
_process_node_cite(tree)
s = html.tostring(tree, encoding='unicode')
if s.endswith('</div>'):
s = s[:-6]
if s.startswith('<div>'):
s = s[5:]
return s
def _process_node_cite(node):
"""Do the citation replacement as we walk the lxml tree."""
def _get(o, name):
value = getattr(o, name, None)
return '' if value is None else value
if 'data-cite' in node.attrib:
cite = '\cite{%(ref)s}' % {'ref': node.attrib['data-cite']}
prev = node.getprevious()
if prev is not None:
prev.tail = _get(prev, 'tail') + cite + _get(node, 'tail')
else:
parent = node.getparent()
if parent is not None:
parent.text = _get(parent, 'text') + cite + _get(node, 'tail')
try:
node.getparent().remove(node)
except AttributeError:
pass
else:
for child in node:
_process_node_cite(child)